Risk Warning of Power Grid Congestion Events Based on Probabilistic Power Flow

To address power grid congestion events stemming from the uncertainty of wind power output, this paper presents a risk early warning model for power grid congestion events based on probability power flow. Employing the deep mixture density network algorithm to compute probability power flow (PPF) enables accurate early warning of power system congestion events. Experimental results conducted on the IEEE 118 demonstrate that the proposed early warning method exhibits high prediction accuracy, requires minimal time consumption, and demonstrates strong generalization. The findings of this study contribute to enabling the power grid to make more proactive dispatching decisions.

I. Introduction

Power System. The random fluctuation of new energy generation significantly increases the uncertainty of power grid operation, resulting in substantial discrepancies between day-ahead and intra-day operation modes. This discrepancy exacerbates the challenge of avoiding power grid congestion events during the real-time scheduling stage. The current scheduling system operates on a 15-minute rolling basis, officially clearing the real-time market for the subsequent two hours. However, this model lacks sufficient forward-looking capabilities to anticipate future uncertain scenarios, leading to a lack of systematic early warning and control mechanisms for power grid congestion events caused by new energy uncertainty. Consequently, it is susceptible to the issue of multi-resource collaborative real-time scheduling, characterized by 'no peak supply, no valley use, and no slope climbing,' which causes scheduling strategy formulation to lag and be overly reliant on past experiences. Hence, it is imperative to provide advance warnings of power grid congestion events, furnish guidance information for subsequent scheduling decisions, and allocate adequate decision-making time to effectively mitigate future security risks.